Adaptive processor convergence improvement using reiterative projection statistics

Adaptive radar processors form estimates of the statistics of the received interference (such as clutter and/or jamming) and receiver noise processes using measured samples (i.e., snapshots) of the signal environment. Snapshots that contain the signal of interest (i.e., targets) and/or other outliers are, in practice, frequently interspersed within a set of more homogeneous interference snapshots. This condition often results in poor convergence in terms of signal to interference-plus-noise ratio (SINR) and ultimately, probability of detection. In this paper, a previously developed projection statistics (PS)-based outlier detection technique is extended to a reiterative and prewhitened form, similar to a recent reiterative generalized inner product (GIP) technique. We compare SINR convergence performance of reiterative GIP and reiterative prewhitened PS, among other methods, in the presence of multiple outliers. The results show that reiterative prewhitened PS is superior to reiterative GIP and to the other methods in terms of SINR convergence criteria.